It seems that mild goto usage is not uncommon in at least some major golang projects:

https://github.com/dotcloud/docker/search?q=goto&ref=cmdform

https://github.com/hashicorp/serf/search?q=goto&ref=cmdform

https://github.com/skynetservices/skydns/search?q=goto&ref=c...

https://github.com/calmh/syncthing/search?q=goto&ref=cmdform

https://github.com/juju/juju/search?q=goto&ref=cmdform

EDIT: formatting

Why? To be sure they should be used sparingly, but it looks it makes enough sense in this particular case and isn't single-handedly going to cause any spaghetti-code problems.

In this particular case I think a goto fits the bill.

Go is fairly strict in the use of goto: http://golang.org/ref/spec#Goto_statements

They have their use cases, and this code seems like a fine use of it. Goto is taught as one of those things you should never use but it still has a purpose. If you look at any C code goto still used. You just have to make sure not to overuse it.

At least in this situation, you can see the entire affected control flow within four lines.

goto generally becomes a problem when suddenly everything is short-circuiting everywhere and you can't see where control's getting passed off to.

All tools can be misused. Goto has a dark past, but it can still be used responsibly. It's almost like generalizing a concept like OOP to be bad because horrible code can be made with it.

Sure. This is a fair use case. Kind of makes more sense than a full while loop, especially since the case is only likely to occur at most once per call to GetRandomName().

If you use it responsibly, I don't see much problem.

or a loop that is not guaranteed to end?

PRNGs don't work that way. They have periods over which all possible output values are produced, close to an equal number of times. Even when producing tuples, we know that this won't loop forever.

You know that, because you already know how "rand" is implemented